Typical performance analysis, debugging, and similar tools (hereinafter referred to collectively as “tracing tools” and “monitoring” tools) help a user identify problems by, e.g., recording information about one aspect of the behavior of a process. Such tools often collect a voluminous amount of information, including information that is unrelated to helping detect the anomaly or anomalies being sought by the user. In addition to making it more difficult to locate the cause of a problem, the excessive amount of information collected and displayed to the user typically comes at a performance cost due to the resources consumed by such tracing tools.
Therefore, it would be desirable to have a better way to identify anomalies.